The focus of the analysis is the Clinical Resource Hub program, a regionally organized initiative designed to address one of the most persistent operational headaches in large health systems: temporary staffing gaps. Under the program, each of the VHA’s eighteen administrative regions, known as Veterans Integrated Service Networks, hosts a hub that deploys clinical staff to fill vacancies at ambulatory care sites, or spokes, throughout the region. Because each hub operates within a region with its own veteran population, facility mix, and workforce conditions, the program’s outcomes vary considerably across the country. Understanding whether those variations follow recognizable regional patterns could help leaders design and evaluate the program more effectively.

The research team set out to test whether hierarchical agglomerative clustering, a classic unsupervised learning method, could accomplish two things: reveal interpretable regional typologies, and demonstrate face validity for learning about differences in Clinical Resource Hub outcomes. The choice of method matters. Many clustering approaches used in health services research produce groupings that are statistically valid but difficult for operational leaders to interpret or act upon. The authors wanted a technique whose output would align with the intuitions of people who actually manage regional programs, making the results usable rather than merely academic.

The methodological pipeline began with feature selection. The team developed a set of regional characteristics with hypothesized links to variations in the Clinical Resource Hub program, drawing on veteran demographics, health status, and facility and staffing variables already defined in existing VHA data. Veteran-level data were aggregated to the regional level, producing sixteen features that described each region’s population and infrastructure. These drew on the VHA’s Corporate Data Warehouse, the electronic health record, and the Survey of Healthcare Experience of Patients, alongside measures such as the Care Assessment Needs score, which captures patient clinical complexity.

With sixteen standardized regional features in hand, the researchers applied principal component analysis to reduce dimensionality. PCA transforms correlated variables into a smaller set of uncorrelated principal components, each a weighted combination of the original features. By examining the feature loadings, the weights showing how strongly each original variable contributes to a component, the team could interpret what each component represented, for example a dimension capturing rurality and age versus one capturing growth and demographic change. This interpretive step is what separates the study’s approach from black-box clustering: rather than feeding raw variables into an algorithm, the researchers first distilled them into comprehensible axes of regional difference.

Hierarchical agglomerative clustering was then applied to the retained principal components. In HAC, each observation, here a VHA region, begins as its own cluster, and the algorithm iteratively merges the most similar clusters according to a linkage criterion, building a hierarchy from the bottom up. The researchers iteratively tested the face validity of the resulting groupings, checking whether the clusters made sense in light of operational knowledge. The procedure settled on four clusters, four distinct regional profiles. One profile captured regions that were older, sicker, and more rural; another described regions that were younger, growing, and with a larger share of female veterans. The other two profiles occupied intermediate or contrasting positions along these dimensions.

The payoff came when the team compared program outcomes across these clusters. They examined access before and after the Clinical Resource Hub program, using fiscal year 2019 as the pre-program baseline and fiscal year 2023 as the post-program period, and utilization in fiscal year 2023. Apparent differences emerged between clusters. Notably, one cluster displayed both the lowest variation in access scores and the lowest absolute access score, a combination the authors interpret as signaling regional access challenges: consistently low performance across the regions in that group rather than a mix of strong and weak performers. Such a pattern is exactly the kind of signal that could prompt targeted attention from program leadership.

Importantly, the regional typologies produced by the PCA-informed clustering aligned with insights from operational leadership, suggesting the method achieved the interpretability the team sought. This alignment between statistical output and managerial intuition is a meaningful validation for a proof-of-concept study. It indicates that the clustering was not detecting mathematical artifacts but was capturing real, recognizable differences in the populations and conditions that each regional hub serves. For a health system contemplating how to evaluate regionally based programs, that is a promising demonstration that unsupervised learning can support, rather than obscure, practical decision-making.

The study also carries methodological lessons for the broader field of health services research. Large integrated systems increasingly regionalize services to optimize resource allocation and improve efficiency, yet evaluations of such programs often treat regions as interchangeable units or rely on ad hoc groupings. The approach demonstrated here, aggregating patient-level data to regional features, reducing dimensionality with PCA, and clustering with HAC, offers a transparent, reproducible template that other systems could adapt. The authors followed established reporting standards, using the STROBE guidelines for observational studies and the SQUIRE guidelines for quality improvement work, and the evaluation was conducted as part of an ongoing VHA quality improvement effort rather than human subjects research.

The authors are careful to frame the work as a first step rather than a finished solution. They note that future work is needed to confirm the usefulness of this method more generally as an approach for understanding outcome variations across regionally based programs. The study was funded by the Department of Veterans Affairs Veterans Health Administration through the VA Office of Rural Health and the VA Office of Primary Care, with the funders having no role in the design, analysis, or decision to publish. Published open access on 7 October 2026, the article arrives at a moment when health systems everywhere are grappling with workforce shortages and uneven access. If simple, interpretable clustering can help leaders see which regions face structural headwinds, programs like the Clinical Resource Hub can be tuned to meet veterans where they actually are, rather than where an average assumes they are.

Subject of Research: Unsupervised clustering of regional characteristics in the Veterans Health Administration's Clinical Resource Hub program

Article Title: Understanding regional variations in the veterans health administration’s clinical resource hub program using hierarchical agglomerative clustering

Article References: Mayfield, B., Wheat, C., Stockdale, S. E., Rose, D. E., Kath, S., Ford, G., Curtis, I., Nelson, K. M., & Rubenstein, L. V. (2026). Understanding regional variations in the veterans health administration’s clinical resource hub program using hierarchical agglomerative clustering. BMC Health Services Research. https://doi.org/10.1186/s12913-026-15781-8

Image Credits: AI Generated

DOI: 10.1186/s12913-026-15781-8

Keywords: Veterans Health Administration, Clinical Resource Hub, hierarchical agglomerative clustering, principal component analysis, unsupervised machine learning, health services research, regional variation, access to care, program evaluation, veterans health, dimensionality reduction, staffing gaps

The research, published in the Journal of General Internal Medicine, analyzed an extraordinary dataset: 1.6 million Veterans who made at least one emergency department visit between October 2022 and September 2023, generating 3.2 million individual visits. Of those Veterans, 185,392 were women, accounting for 362,347 emergency visits. The investigators, led by Dr. Jessica Faiz of the VA Greater Los Angeles Healthcare System and UCLA, drew on VA administrative records, community care claims data, and enrollment files to reconstruct where every visit occurred and why. Their central finding is deceptively simple: about 60 percent of women Veterans’ emergency visits took place in VA emergency departments, while the remaining 40 percent occurred in community settings paid for through VA Community Care, the program expanded by the MISSION Act of 2018 that allows Veterans to receive care outside the VA system.

That 60-40 split matters because the two populations flowing through each door are strikingly different. Women Veterans who used VA emergency departments were, on average, younger than their male counterparts, with a mean age of 50.2 years compared with 62.9 years for men. They were also more likely to be Black, representing 33 percent of women emergency users compared with 20.9 percent of men, more likely to be Hispanic, more likely to live in urban areas, and more likely to carry a service-connected disability, at 65.5 percent versus 50.1 percent. Most striking of all, 42.1 percent of women Veterans presenting to emergency departments had a serious mental illness, nearly double the 23.4 percent prevalence among men. The picture that emerges is of a high-need population whose emergencies are intertwined with mental health, housing instability, and the long shadow of military service.

To untangle why some women end up in VA emergency departments and others in community ones, the team deployed logistic regression with generalized estimating equations, a statistical technique that accounts for the fact that the same patient may appear many times in the dataset across repeated visits. The models adjusted for age, race, ethnicity, rurality, comorbidity burden measured by the Elixhauser index across 31 chronic conditions, serious mental illness, housing instability, combat exposure, military service branch, VA enrollment priority group, driving time to the nearest VA emergency department or urgent care clinic, and the frequency of primary care contact in the preceding year. The result is one of the most comprehensive adjusted portraits of emergency care navigation ever assembled for this population.

The diagnosis a woman carries through the door turns out to be one of the strongest predictors of where she will be treated. Women with low-acuity conditions, things like upper respiratory infections, sinusitis, low back pain, and skin disorders, had 62 percent higher odds of being treated in a VA emergency department than in the community. Behavioral health-related visits also tilted toward VA settings, with 10 percent higher odds. The pattern likely reflects the financial predictability of VA care: for non-urgent problems, Veterans may reasonably prefer a system where copays are transparent and coverage is guaranteed, rather than risk surprise bills or gaps in Community Care authorization. In contrast, women presenting with emergency care-sensitive conditions, the time-critical diagnoses such as myocardial infarction and stroke where rapid, high-quality treatment demonstrably reduces death and disability, had 39 percent lower odds of receiving care at a VA facility. Chest pain was the single most common diagnosis among women treated in community emergency departments, while musculoskeletal pain topped the list at VA sites.

Geography exerts an equally powerful force. Women who lived farther from a VA emergency department were substantially more likely to seek care in the community, a finding that aligns precisely with the intent of the MISSION Act, which was designed to relieve Veterans facing geographic barriers to VA facilities. Distance also appears to shape the behavior of women experiencing homelessness, who had 10 percent lower odds of using VA emergency care. The authors suggest that fragmented care and lost continuity with VA services, chronic challenges for homeless-experienced Veterans, may push this group toward whatever emergency department is closest, regardless of affiliation. Notably, women with pregnancy-related diagnoses had the lowest odds of any group of presenting to a VA emergency department, consistent with the fact that most women Veterans receive obstetric care outside the VA system entirely.

Perhaps the most socially consequential finding concerns race and trust. Women Veterans from racially and ethnically minoritized groups had consistently higher odds of using VA emergency departments than White women, even after adjusting for clinical and geographic factors. The researchers offer two non-exclusive explanations: this may signal genuine trust in VA providers, a phenomenon documented in prior studies of women Veterans’ healthcare experiences, or it may reflect structural barriers, cost, insurance status, provider availability, that make community care harder to reach. Either way, the pattern echoes earlier national data showing that minoritized Veterans rely more heavily on VA-provided care across the board. It also raises the stakes for equity in emergency settings, where split-second clinical decisions intersect with providers who, as the broader literature on implicit bias in emergency medicine shows, are not immune to the cognitive shortcuts that can shape outcomes.

Combat exposure emerged as another unexpected predictor. Women Veterans with a history of military combat had 20 percent higher odds of seeking emergency care at VA facilities. The authors speculate that this may reflect the comprehensive post-combat care the VA offers, a stronger institutional connection among combat Veterans, or the higher rates of service-connected disability that qualify them for lower copays. Whatever the mechanism, the finding reinforces a theme running through the entire study: the VA emergency department functions not merely as a treatment site but as an anchor of institutional identity and trust for a subset of women whose needs span trauma, mental health, and physical medicine simultaneously.

The study’s technical rigor is matched by candid acknowledgment of its limits. Administrative data cannot capture clinical nuance, urgency, or patient preference, all of which influence where someone chooses, or is able, to seek care. The researchers could not distinguish ambulance arrivals from walk-ins, a crucial gap because community emergency department use for chest pain or stroke may reflect emergency medical services routing protocols rather than any deliberate choice. The analysis also captured only VA-paid visits, missing care covered by private insurance or Medicare. Missing race and ethnicity data, affecting 8.2 percent of women, were handled through multiple imputation by chained equations, and multicollinearity among predictors was formally assessed using variance inflation factors, all of which fell below 1.7 after collinear variables were removed.

The implications reach far beyond the VA. Because two in five women Veterans’ emergency visits now occur in community settings, and because those community visits are disproportionately the high-acuity, high-admission encounters most likely to require follow-up, the authors argue that care coordination between systems is no longer optional. They call for efficient exchange of emergency records, medication changes, and diagnostic results between community emergency departments and VA primary and specialty care; routine ascertainment of Veteran status at community registration desks; and automated notifications to VA care teams when a Veteran presents in the community. Within the VA itself, they point to the need for expanded behavioral health capacity, round-the-clock ultrasound and gynecologic consultation, and same-day services for lower-acuity concerns. As the population of women using VA healthcare has nearly doubled over the past decade, the study makes clear that the question is no longer whether women Veterans will receive emergency care in both systems, but whether both systems will be ready for them.

Subject of Research: Emergency department utilization patterns among women Veterans across VA and community care settings

Article Title: Where Women Veterans Seek Emergency Care: Patterns and Predictors of VA and Community Emergency Department Use

Article References: Faiz, J., Carvalho, C. J., Urech, T., Wu, S., Breland, J. Y., Frayne, S. M., Washington, D. L., Cordasco, K. M., & Vashi, A. A. (2026). Where Women Veterans Seek Emergency Care: Patterns and Predictors of VA and Community Emergency Department Use. Journal of General Internal Medicine. https://doi.org/10.1007/s11606-026-10847-z

Image Credits: AI Generated

DOI: 10.1007/s11606-026-10847-z

Keywords: women Veterans, emergency department, Veterans Health Administration, VA Community Care, MISSION Act, behavioral health, health equity, care coordination, geographic access, homelessness, mental health, health services research

At the heart of the exchange is a question that has long troubled sleep researchers: how do you study mortality in a disease as rare as narcolepsy without waiting decades for a prospective cohort to mature? Narcolepsy affects roughly one in two thousand people, and its two main forms, narcolepsy type 1 with cataplexy and narcolepsy type 2 without cataplexy, are distinguished by features that require specialized testing to confirm. The authors argue that for rare conditions, electronic health record based cohort studies remain indispensable precisely because the prospective designs their critics favor are seldom feasible at the scale or duration needed to detect mortality signals. Their study leveraged the Veterans Health Administration’s longitudinal records, which follow patients across decades of care, offering a follow-up window of 25 years that no single-center prospective study could realistically match.

One of the sharpest criticisms concerned bias. Mehri and Finsterer invoked limitations commonly associated with retrospective designs, including recall bias, selection bias, and confounding. The authors’ reply draws a technical distinction here that is worth unpacking. Recall bias, the distortion that arises when patients or investigators remember exposures imperfectly, is a meaningful threat in studies that rely on interviews or chart reviews conducted after the fact. It is not, however, a meaningful threat in administrative electronic health record cohorts, where exposures and outcomes are recorded prospectively at the point of care, independent of anyone’s memory. A diagnosis code entered during a clinic visit and a death recorded in a vital status file are not subject to the fallibilities of recollection. Selection bias and confounding, by contrast, the authors concede, remain legitimate concerns in any observational dataset, and they describe the specific countermeasures they deployed.

Those countermeasures centered on propensity score matching, a statistical technique widely used in non-randomized research to approximate the balance achieved in randomized trials. In the narcolepsy study, patients with narcolepsy were matched to comparator patients from general sleep clinics on age, sex, race and ethnicity, and index diagnosis year, with additional adjustment for body mass index and the Charlson Comorbidity Index, a validated composite measure of baseline disease burden. The authors also align themselves with earlier correspondents, Talari and Goyal, who noted that retrospective designs cannot establish causality. Their own manuscript, they emphasize, explicitly frames the findings as associations rather than causal effects. This is a crucial framing for readers: the study does not claim that narcolepsy itself causes death, but that in real-world clinical practice, patients diagnosed and managed for narcolepsy died at higher rates than comparable sleep clinic patients, even when those comparators carried a substantially heavier comorbidity burden at baseline.

The reliability of the mortality outcome itself became another point of contention, and here the authors marshal quantitative evidence. Vital status data in the Veterans Health Administration have been independently validated against the National Death Index, the United States’ authoritative repository of death records. That validation work, published by Sohn and colleagues, found a sensitivity of 98.3 percent and exact-date agreement of 97.6 percent, figures that support the accuracy and completeness of the VHA’s mortality ascertainment. In practical terms, this means the study’s primary endpoint, all-cause death, was measured against a benchmark that few national datasets can match. The authors argue that these psychometric properties of the data source underpin the reliability of their central finding, whatever reservations may attach to the diagnostic classification of the exposure.

That diagnostic classification was the critics’ most technically substantive objection. Definitive diagnosis of narcolepsy type 1 under the International Classification of Sleep Disorders, third edition, requires objective confirmation: polysomnography followed by a multiple sleep latency test demonstrating shortened sleep latency and early REM onset, or cerebrospinal fluid measurement revealing low hypocretin-1, the neuropeptide whose loss underlies the disorder’s pathophysiology. The critics asked how many participants had cataplexy and whether all had undergone these tests. The authors’ answer is a candid acknowledgment of the limits of administrative data: polysomnography and multiple sleep latency test results are not captured in a standardized format across the national VHA dataset, and cerebrospinal fluid hypocretin testing is rarely performed in routine Veterans Affairs practice. These data simply cannot be ascertained reliably from administrative records, a constraint that shaped the study’s entire design philosophy.

Rather than attempting an unattainable gold-standard confirmation, the researchers employed a computable phenotyping algorithm, requiring at least two narcolepsy diagnosis codes separated by 30 to 390 days. This temporal separation requirement is designed to improve diagnostic specificity by filtering out random or provisional miscoding, since a genuine narcolepsy diagnosis tends to persist across multiple encounters while a coding error is less likely to be repeated weeks or months later. The authors note that this approach aligns with prior large electronic health record based studies of narcolepsy. They also push back on the implicit assumption that the diagnostic gold standard is itself flawless, citing work by Trotti and colleagues showing that approximately half of patients with noncataplectic central hypersomnia received a different diagnosis when the multiple sleep latency test was repeated. Even in tertiary referral settings, the boundary between narcolepsy type 2 and idiopathic hypersomnia is unstable, which complicates any demand for perfect phenotypic precision.

The critics also raised the specter of confounding by conditions and medications that mimic or modify hypersomnolence: endocrine disorders such as hypothyroidism, and drug classes including antiepileptics, analgesics, antihistamines, and immunosuppressants. The authors offer a reasoned rebuttal on both fronts. Hypothyroidism and similar conditions, they argue, would more plausibly be misclassified within the general sleep clinic comparator group than within the narcolepsy cohort, where the two-code requirement reduces the likelihood of diagnostic confusion. As for the medication classes, none has a well-established direct effect on the mortality pathways most likely operative in narcolepsy, namely cardiovascular events and accident-related deaths, and adding them would have introduced collinearity without addressing the principal causal hypotheses. The study instead focused on comorbidity groups with established relevance to mortality risk in narcolepsy: cardiovascular, metabolic, pulmonary, neurologic, and psychiatric conditions. The authors caution that expanding the model to every conceivable confounder would reduce interpretability, destabilize the statistical models, and exceed the constraints of available structured data.

The exchange is a window into a broader tension in modern epidemiology. Big data cohorts built from electronic health records offer scale, duration, and real-world generalizability that carefully phenotyped prospective studies cannot match, but they sacrifice diagnostic granularity in return. The authors of the reply do not dispute that prospectively defined cohorts are essential for answering mechanistic questions and refining diagnostic precision; they simply note that such cohorts are logistically difficult to obtain for rare disorders and cannot provide the scale or quarter-century of follow-up available in the VHA population. Their study, they reiterate, was never designed to establish causality or to revalidate diagnostic criteria, but to assess whether narcolepsy, as diagnosed and managed in ordinary clinical practice, is associated with increased mortality and healthcare utilization. On that question, they argue, the answer stands: the elevated mortality persisted even against a comparator group with a higher baseline comorbidity burden, reinforcing the clinical relevance of the finding and the case that real-world evidence, carefully constructed, deserves a seat at the table in rare-disease research.

Subject of Research: Mortality risk and study methodology in narcolepsy research using large electronic health record cohorts

Article Title: Reply to “The question of whether narcolepsy increases mortality should be answered based on reliable studies”

Article References: Sharafkhaneh, A., BaHammam, A. S., Dauvilliers, Y., Azarian, M., Thorpy, M., Han, F., Senel, G. B., Aksu, M., & Razjouyan, J. (2026). Reply to “The question of whether narcolepsy increases mortality should be answered based on reliable studies”. Journal of Clinical Sleep Medicine, 22(1), Article 110. https://doi.org/10.1007/s44470-026-00121-8

Image Credits: AI Generated

DOI: 10.1007/s44470-026-00121-8

Keywords: narcolepsy, mortality, electronic health records, retrospective cohort study, propensity score matching, Veterans Health Administration, sleep medicine, computable phenotyping, polysomnography, multiple sleep latency test, confounding, hypocretin

The research team, led by Zachary L. Mannes of Columbia University Irving Medical Center and Deborah S. Hasin of Columbia University and the New York State Psychiatric Institute, focused on VHA patients aged 18 to 75 with chronic pain diagnoses. Veterans are a population of particular interest in this debate because chronic pain is among the most common conditions treated in the VA system, and the department has spent more than a decade working to reduce opioid exposure among its enrollees. The VA reports having cut the number of veterans prescribed opioids dramatically since 2012, through initiatives such as the Opioid Safety Initiative and updated clinical practice guidelines that emphasize non-opioid pain management. Against that backdrop, any independent effect of state cannabis policy is difficult to detect, which makes the new study’s methods especially important.

To isolate the effect of legalization from these broader secular trends, the investigators used staggered-adoption difference-in-difference models, a technique that has become the workhorse of modern policy evaluation. Because different states legalized medical cannabis and recreational cannabis at different times, the researchers could compare prescribing outcomes within the same patients and states before and after each law took effect, using states without such laws as the comparison group. States were categorized each year as having no medical or recreational cannabis law, a medical cannabis law only, or both. This approach controls for fixed characteristics of states and for national trends in opioid prescribing, such as the prescribing declines that followed the 2016 CDC opioid guideline, leaving the change associated with legalization itself.

The outcomes the team tracked were carefully chosen to capture different dimensions of opioid risk. They examined whether patients received at least 30 days of prescription opioids, whether they were on long-term opioid therapy, defined as at least 90 consecutive days of opioid supply without a gap of more than 15 days, whether they received high-dose opioids of at least 50 morphine milligram equivalents per day, and whether opioids were co-prescribed with benzodiazepines, a combination strongly linked to overdose death. Each of these measures reflects a distinct clinical decision point, from initiating an opioid prescription to escalating dose or layering on a sedative, and each has been flagged in national guidelines as a marker of potentially hazardous prescribing.

The results paint a nuanced picture. Enactment of a medical cannabis law was associated with a reduction of 0.79 percentage points in the share of patients receiving at least 30 days of opioids, and a reduction of 0.34 percentage points in long-term opioid therapy, relative to states without such laws. Those numbers may sound small, but applied across the enormous VHA population with chronic pain, they translate into thousands of patients shifting away from extended opioid courses. Notably, however, medical cannabis laws were associated with slight increases in the other two risk markers: high-dose opioid prescribing rose by 0.09 percentage points and opioid-benzodiazepine co-prescribing rose by 0.11 percentage points, changes the authors interpret cautiously given their small magnitude.

The story changed when recreational cannabis laws entered the picture. Following the enactment of recreational legalization, on top of existing medical laws, the researchers observed further declines in at least 30-day opioid prescribing, with an additional reduction of 0.14 percentage points, and more substantial drops in the two high-risk markers. High-dose opioid prescribing fell by 0.33 percentage points and opioid-benzodiazepine co-prescribing fell by 0.33 percentage points compared with the period after medical legalization alone. In other words, the transition from medical-only to full legalization was associated not just with less opioid prescribing overall, but with less of the kind of prescribing most closely tied to overdose risk, a pattern that runs counter to fears that wider cannabis availability would compound dangerous drug combinations.

Age emerged as a critical modifier of these effects. The associations between cannabis laws and opioid prescribing were generally more consistent among adults aged 65 to 75, the oldest group in the cohort. This finding is striking given that cannabis use has been rising fastest among older Americans in recent years, and given that older adults face elevated risks from both opioids and benzodiazepines due to altered drug metabolism, falls, and respiratory depression. If legal cannabis access is substituting for opioids anywhere, the data suggest it may be among older veterans with chronic pain, a group in whom clinicians may be particularly motivated to taper high-risk regimens when an alternative becomes legally available.

The authors are careful about what these associations do and do not mean. Difference-in-difference estimates capture population-level changes in prescribing, not individual decisions, and the study cannot confirm that any specific patient replaced an opioid with cannabis. The plausible mechanism, the researchers suggest, involves changes in both patient and clinician treatment decisions: patients in legal states may request cannabis as part of their pain management, and clinicians may feel more comfortable tapering opioids when patients have a sanctioned alternative. Survey evidence has long shown that chronic pain is the leading qualifying condition for medical cannabis licenses, and prior studies in Medicare and Medicaid populations have reported similar directional findings, though the literature overall has been mixed, with some studies finding no effect or effects concentrated in specific subgroups.

The study also arrives amid ongoing scientific uncertainty about whether cannabis actually relieves chronic pain. Randomized trials and meta-analyses of cannabinoids for chronic non-cancer pain have produced modest and inconsistent evidence of benefit, and recent best-practice advice from the American College of Physicians acknowledges both the limited efficacy data and the potential harms of cannabis use. Previous work by some of the same researchers found that cannabis legalization was associated with increases in cannabis use disorder diagnoses among VHA patients with chronic pain, a reminder that expanded access carries its own risks. The new findings therefore do not settle whether cannabis is an effective analgesic; they document a shift in prescribing behavior that accompanies legalization, whatever its underlying cause.

What the study does offer is the most comprehensive picture to date of how cannabis policy intersects with opioid care in the largest integrated health system in the United States. By following VHA patients through a decade in which the legal landscape transformed state by state, and by separating medical from recreational laws and dissecting prescribing into clinically meaningful risk categories, the research moves the debate beyond anecdote. The message for policymakers is measured: greater legal cannabis access appears associated with modest reductions in opioid prescriptions for chronic pain, particularly long-term therapy and, after recreational legalization, high-risk prescribing patterns, while the clinical question of whether cannabis should be recommended for pain remains open. As more states reconsider their cannabis laws, studies like this one will be essential for weighing the trade-offs embedded in those decisions.

Subject of Research: Association between state cannabis legalization and opioid prescribing patterns among veterans with chronic pain

Article Title: Cannabis Legalization and Opioid Prescribing in Veterans Health Administration Patients: 2013–2022

Article References: Mannes, Z. L., Wall, M. M., Stohl, M., Malte, C. A., Olfson, M., Livne, O., Fink, D. S., Simpson, T., Wisell, C. G., Keyhani, S., Martins, S. S., Cerdá, M., Sacco, D. L., Gutkind, S., Maynard, C. C., Sherman, S., Saxon, A. J., & Hasin, D. S. (2026). Cannabis Legalization and Opioid Prescribing in Veterans Health Administration Patients: 2013–2022. Journal of General Internal Medicine. https://doi.org/10.1007/s11606-026-10733-8

Image Credits: AI Generated

DOI: 10.1007/s11606-026-10733-8

Keywords: cannabis legalization, opioid prescribing, chronic pain, Veterans Health Administration, medical cannabis law, recreational cannabis law, long-term opioid therapy, benzodiazepine co-prescribing, difference-in-difference, older adults, pain management, public health policy

The research team, led by investigators including Ramin Ebrahimi, Paul A. Dennis, and Jennifer A. Sumner, drew on national electronic health records spanning 2000 to 2019, a period covering multiple military service eras. Women veterans represent the fastest-growing group within the Veterans Health Administration, nearly tripling in number from roughly 160,000 in 2000 to nearly 440,000 in 2015, and they are projected to account for nearly 18 percent of all living US veterans by 2040. Yet despite carrying a heavier burden of both psychiatric and traditional cardiovascular risk factors than civilian women or male veterans—and experiencing higher cardiovascular mortality—they have been strikingly underrepresented in research on the PTSD-heart disease connection.

Methodologically, the study was designed to guard against the confounding that has plagued earlier work in this field. The investigators identified 622,312 women veterans with at least one VHA healthcare encounter, then excluded anyone with pre-existing cardiovascular disease. From this pool, 140,210 women with documented PTSD were matched to 101,448 women without the disorder using propensity score methods. Rather than relying on conventional statistical matching, the team employed extreme gradient boosting, a machine learning algorithm that iteratively models residuals to capture complex interactions among dozens of variables, including demographics, traditional risk factors such as hypertension and smoking, female-specific risks like preeclampsia, and a range of mental health conditions from depression to substance use disorders.

The matching process achieved excellent balance across all measured covariates, with standardized mean differences falling below the conventional threshold of 0.10. The resulting cohort had a mean age of 39.1 years, with fewer than 5 percent aged 60 or older, and was notably diverse: just over half were non-Hispanic White, roughly 30 percent non-Hispanic Black, and more than 9 percent Hispanic/Latina. Cardiovascular outcomes were identified using strict diagnostic code criteria—requiring at least one inpatient or two outpatient records for each condition—an approach validated in prior Department of Veterans Affairs research with reported sensitivity and specificity above 0.95 for PTSD identification.

Over an average follow-up of roughly six years, 7.5 percent of women with PTSD developed at least one of the cardiovascular conditions in the composite outcome, compared with 5.5 percent of women without PTSD—a 36 percent relative increase in crude cumulative incidence. In survival models accounting for the competing risk of death, PTSD was associated with a 24 percent higher rate of incident cardiovascular disease overall. The association held for ischemic heart disease, with a hazard ratio of 1.30, and for stroke and transient ischemic attack, with a hazard ratio of 1.29, consistent with meta-analytic estimates from earlier studies that had reported pooled hazard ratios of approximately 1.6 for coronary heart disease and stroke.

What distinguishes this study is its reach beyond the familiar territory of heart attacks and strokes. The researchers found significant associations between PTSD and heart failure, venous thromboembolism, pulmonary hypertension, and aortic valve stenosis. The link with venous thromboembolism was among the strongest observed, with a hazard ratio of 1.34, while aortic stenosis showed a hazard ratio of 1.23 and pulmonary hypertension 1.17. Associations with atrial fibrillation and peripheral arterial disease were elevated but did not reach statistical significance, possibly because these conditions remain rare in a cohort of women in their late thirties followed for only about six years.

Perhaps the most striking findings emerged from stratified analyses. The PTSD-cardiovascular connection was strongest in the youngest women: those aged 30 to 39 showed a hazard ratio of 1.52, and those aged 18 to 29 a hazard ratio of 1.42, compared with more modest elevations in older age groups. Racial and ethnic disparities were equally pronounced. American Indian/Alaska Native women with PTSD had a 58 percent higher rate of cardiovascular disease, and Hispanic/Latina women a 50 percent higher rate, compared with their counterparts without PTSD. The authors caution that small sample sizes for some subgroups produced wide confidence intervals, but the pattern suggests that certain populations may bear a disproportionate share of PTSD-related cardiovascular risk.

The biological plausibility of these associations rests on well-characterized pathways. PTSD has been linked to unhealthy behaviors including physical inactivity, poor diet, disrupted sleep, and smoking, as well as to comorbid depression and substance use disorders that independently elevate cardiovascular risk. Beyond behavior, the disorder involves dysregulation of the hypothalamic-pituitary-adrenal axis and hyperreactivity of the sympathetic nervous system, particularly in response to trauma-related reminders. These stress-system disturbances can drive vasoconstriction, endothelial dysfunction, arterial stiffness, and increased cardiac afterload—mechanisms that plausibly contribute to ischemic heart disease and stroke. Immune dysregulation and chronic systemic inflammation, also documented in PTSD, may underlie connections to heart failure and pulmonary hypertension specifically.

Notably, prior work by the same research group found that traditional and nontraditional risk factors together explained less than half of the association between PTSD and incident ischemic heart disease in women veterans, implying that much of the excess risk operates through pathways not yet captured in electronic health records. The authors call for mechanism-focused research into modifiable factors such as sleep patterns, physical activity, and trauma-related processes, which could open new avenues for prevention tailored to this high-risk population.

The clinical implications are difficult to ignore. Current cardiology guidelines have historically categorized women under 45 as low-risk for cardiovascular events, and the most recent VA/Department of Defense dyslipidemia guidelines recommend screening only from age 40, with no mention of PTSD status. Yet this study found the strongest PTSD-related risk precisely in women under 40, and earlier research showed that ten-year cardiovascular risk rises steadily from age 30 in women veterans. The findings argue for earlier and more tailored cardiovascular risk assessment in women veterans with PTSD, and the Veterans Health Administration has begun developing a Cardiovascular Toolkit aimed at improving risk identification and patient education for women veterans. The researchers also emphasize the need to determine whether similar early risk elevation exists in non-veteran women and men with PTSD, to verify the novel associations with aortic stenosis and pulmonary hypertension, and to test whether targeted interventions can blunt the cardiovascular toll of trauma-related illness.

Subject of Research: The association between posttraumatic stress disorder and incident cardiovascular disease in women veterans

Article Title: Impact of posttraumatic stress disorder on incident cardiovascular outcomes in women veterans: a retrospective longitudinal study

Article References: Ebrahimi, R., Dennis, P. A., Alvarez, C. A., Shroyer, A. L., Beckham, J. C., & Sumner, J. A. (2026). Impact of posttraumatic stress disorder on incident cardiovascular outcomes in women veterans: a retrospective longitudinal study. eClinicalMedicine, 100, Article 104197. https://doi.org/10.1016/j.eclinm.2026.104197

Image Credits: AI Generated

DOI: 10.1016/j.eclinm.2026.104197

Keywords: PTSD, cardiovascular disease, women veterans, heart failure, venous thromboembolism, aortic stenosis, pulmonary hypertension, stroke, ischemic heart disease, propensity score matching, Veterans Health Administration, health disparities

Low-value health services—those that provide minimal benefit relative to their costs—pose significant operational and economic challenges within the healthcare system. They can lead to medical overuse, which not only burdens individual patients with unnecessary procedures but also strains system resources, leading to increased healthcare expenditures. The identification of these services is pivotal, as it allows the healthcare system to implement necessary reforms and move towards a more efficient model of care.

Clinicians within the VHA are crucial players in this discourse. They often encounter low-value services in their daily practice, and their insights can illuminate the systemic barriers and facilitators influencing these behaviors. The qualitative study offers a platform for these clinicians to voice their experiences and observations, painting a vivid picture of the systemic complexities at play. It is essential to engage such first-hand accounts, as they can lead to actionable changes in policy and practice.

The research highlights various themes that emerged from clinician interviews. One prominent theme was the tension between patient expectations and the clinical evidence supporting specific treatments or interventions. Clinicians frequently reported instances where patients demanded certain services, driven by misinformation or societal pressures. This dynamic raises questions about the responsibility of healthcare providers to navigate patient desires while advocating for medically sound practices.

Additionally, the study revealed concerns about the influence of external factors on clinician decision-making. One notable factor was the financial incentive structures that may promote low-value care. Clinicians expressed frustration with the alignment of certain reimbursement models that inadvertently encourage the use of non-essential services. This misalignment raises fundamental questions about the long-term sustainability of such healthcare practices and the ethical implications for practitioners working within these constraints.

The findings also pointed to the importance of education and training for both clinicians and patients. Many clinicians advocated for improved educational initiatives that target both groups, aiming to foster understanding about what constitutes low-value care and its implications. Such educational efforts could empower patients to engage more meaningfully in their health decisions, helping them recognize the value of seeking evidence-based care.

Moreover, the emotional and psychological toll on clinicians dealing with low-value service requests was underscored. The study found that the pressure to conform to patient demands could lead to clinician burnout and job dissatisfaction. This issue is particularly concerning as the healthcare workforce faces recruitment and retention challenges, making it imperative to address clinician well-being comprehensively.

The qualitative nature of Cupler et al.’s study allows for an exploration of nuances often overlooked in quantitative research. By leveraging detailed interviews, the researchers captured the complexities of clinician experiences with low-value health services, enhancing the opportunity for understanding and ultimately improving health outcomes for veterans. Such insight is crucial in developing targeted interventions that address not only the services rendered but also the context in which care is delivered.

In light of these findings, it becomes evident that tackling low-value health services requires a multifaceted approach. Collaboration among stakeholders—including policy-makers, healthcare administrators, and clinician groups—is necessary to devise effective solutions. Only through combined efforts can the VHA address the systemic issues that facilitate low-value service use and concurrently improve the overall quality of care veterans receive.

As the study notes, the implications of addressing low-value care extend beyond individual patient interactions. Reducing the prevalence of low-value services can enhance the overall healthcare system’s efficiency, enabling resources to be reallocated to more effective interventions. This not only improves veteran health outcomes but also contributes to broader public health goals.

In summary, Cupler et al.’s qualitative study serves as a vital contribution to the ongoing conversation about low-value health service use within the Veterans Health Administration. By amplifying the voices of clinicians, the research underscores the need for systemic changes that prioritize value-based care. The insights gleaned from this study could help to shift paradigms within the VHA and potentially inform practices across the wider healthcare landscape.

As stakeholders reflect on this research, the call to action is clear: enhancing clinician education, addressing systemic barriers, and fostering a culture that values quality over quantity in healthcare provision must be prioritized. These efforts will ultimately reshape the veteran healthcare experience and set a precedent for the future of healthcare delivery in the United States.

Subject of Research: Clinician perspectives on low-value health service use within the Veterans Health Administration.

Article Title: VA Clinicians’ Perspectives on Low-Value Health Service Use in the Veterans Health Administration: A Qualitative Study.

Article References:

Cupler, Z.A., Hruska, K.L., Beyer, N.M. et al. VA Clinicians’ Perspectives on Low-Value Health Service Use in the Veterans Health Administration: A Qualitative Study. J GEN INTERN MED (2025). https://doi.org/10.1007/s11606-025-09838-3

Image Credits: AI Generated

DOI: 10.1007/s11606-025-09838-3

Keywords: low-value healthcare, Veterans Health Administration, clinician perspectives, qualitative study, healthcare policy, value-based care